On the Contagion of Financial Risk
Özet
The global financial system has become highly interconnected over the past few decades and financial shocks have propagated faster, causing systemic events to occur more frequently. This dissertation examines systemic risk contagion through two linked chapters, each contributing to different strands of the literature. In the first chapter, I construct a contagion test, based on time varying Granger causality and dynamic conditional correlation approaches. Using the test on the systemic risk contributions of international banks, I identify several contagion episodes during the period 2004-2021, particularly concentrated during the four periods of turmoil. I then analyze systemic risk spillovers across international banks following extreme adverse and beneficial shocks, identify the main risk transmitters, and scrutinize changes in network topology during the four contagion episodes. The results reveal that the main transmitters of systemic risk differ not only across magnitudes and directions of shocks, but also across crisis periods. In the second chapter, I investigate the determinants of systemic risk contagion based on tail behavior, taking into account time-variation, slope heterogeneity and endogeneity. Using explanatory variables derived from banks' balance sheets representing size, profitability, capital adequacy, credit quality, leverage, and funding structure I find that determinants of systemic risk contagion change over time, differ in each crisis episode, and no single factor drives contagion persistently. I show that some determinants gradually lose their influence on the propagation of shocks, while others are effective only during a single period of turmoil. The results also show significant heterogeneity across banks, and I do not detect significant clustering at either the national or regional level. The findings reveal that static surveillance methods may fail to capture the factors that propagate systemic risk. In light of my findings, I propose a holistic systemic risk surveillance model that uses high-frequency data and incorporates several risk factors simultaneously.
